Skip to main content
. 2012 Jul 30;28(20):2615–2623. doi: 10.1093/bioinformatics/bts483

Fig. 3.

Fig. 3.

GBM variable importance for functional SNPs. Median GBM variable importance using (a) uncorrected importance, (b) LD subsetting and (c) LD subsetting with PCVs. The skewed distribution of importances makes SEs uninformative, so error bars instead indicate observed upper and lower quartiles. Comparison of the plots shows the reduced impact of LD after LD subsetting and the increased impact of MAF when including PCVs